patrick451

Python3 drops support for older versions of python3 all the time. Every single release comes with a bunch of deprecated and removed features. There is a very strong chance that a python 3.7 codebase is totally broken on 3.14. Almost no language takes backwards compatibility as seriously as c++ does.

Apple not adopting these kinds of user hostile designs is why a lot of us were happy to a premium for their products. I guess Cook is just too stupid to understand that.

"We don't build enough houses" does not explain the massive spike in prices we have seen in the last 5 years.

I would call it it POR-15

Until AVs can deal with OOD scenarios, they should not be on the road.

Most of humanity is mediocre. Very few people are excellent. You're response of "touch luck, just be better" to a population with a mean IQ of 100 will lead to pitchforks in the streets.

In 2010 I paid 3k for a 10 year old truck with 100k miles. That same truck today costs easily 15k. Same story for rent. Same story for groceries. Same story for health insurance.

Who gives a shit how much trinkets costs if you can't afford groceries and rent?

Housing, vehicles, groceries, and health insurance are all up massively. Who gives a shit how much a game costs if you can't afford groceries and rent?

> Could you give some specific examples of AI regulations that you think would be good?

AI companies need to be held liable for the outputs of their models. Giving bad medical advice, buggy code etc should be something they can be sued for.

Imagine if the only way to build a skyscraper was to start with a dollhouse and keep tacking extensions and pieces onto it until. Imagine if the only way to build a bridge across San Francisco bay was to start with pop sickle sticks.

Jumping straight to the new release because it fixed one security bug has always struck me as a round about way of trying to achieve security through obscurity, especially when the releases include tons of other changes. Yes, this release fixed CVE-123, but how many new ones were added?

I wonder how much of this is "I'm saving it for a rainy day" vs "I was conditioned that eating the last one of a treat leads to conflict".

> But field access does not compose easily if you have a nested hierarchy of objects. Your natural choice in the "OOP style" is to write a lot of boiler plate to point to each different field you want to get/set.

This is a self inflicted problem. Make data public and there is no boilerplate.

The conflicts of interest in a lot of AI research is pretty staggering.

Until it goes down because because aws STILL hasn't made themselves completely multi-region or can't figure our their DNS.

Wikipedia says

> A brick (or bricked device) is an electronic device, specially consumer electronics (such as a mobile device, game console, computer, etc.) that is no longer functional.

These jeeps are no longer functional.

> My suspicion is that this was either a CAN saturation issue (ie - infotainment started sending a high priority message which could reach powertrain CAN) or a state management issue (ie - infotainment sent a “put modules to sleep” or “wake modules” message which was not handled correctly and caused one or more modules to transition to an invalid state for driving).

The fact that this possible proves the point: OTA updates are dangerous and should be banned.

The reliability of software is so bad this is an absurd comparison.

> Genuinely curious what the solution here is.

One obvious solution is to ditch drone delivery entirely. It's not a foregone conclusion that we must deliver trinkets by drone.

Seems like a bad decision. Grafana is awful compared to a bespoke solution.

> Either way, I can get arbitrarily good approximations of arbitrary nonlinear differential/difference equations using only linear probabilistic evolution at the cost of a (much) larger state space.

This is impossible. When driven by a sinusoid, a linear system will only ever output a sinusoid with exactly the same frequency but a different amplitude and phase regardless of how many states you give it. A non-linear system can change the frequency or output multiple frequencies.

The problem with the LPs is they are all contradictory. You can make an argument for anything you want using LPs. Want to refactor something? Insist on the highest standards. Don't want to refactor something? Deliver results. It's all BS.

We have a lot of unemployed American tech workers. There is no reason hire people on visas other than suppressing wages and getting what amounts to an indentured servant.

AI makes so many mistakes, I cannot trust it with telling me the truth about how a large codebase works.

What I have noticed is that often, what sleep or rest gives me isn't so much a new insight but just the gumption to make a decision or grind through what I know needs to be done.

> The key advantage of doing batched inference at a huge scale is that once you maximize parallelism and sharding, your model parameters and the memory bandwidth associated with them are essentially free (since at any given moment they're being shared among a huge amount of requests!)

Kind of unrelated, but this comment made me wonder when we will start seeing side channel attacks that force queries to leak into each other.

There are two many obvious exceptions to even start taking this seriously. If we all followed this advice, we would never even multiply matrices.

This has consistently been my experience with every LLM I have tried. Everybody says "Oh, you tried it one the model from two months ago? Doesn't count, the new ones are sooo much better". So I try the new one and it still hallucinates.

> What if you really do need "unsafe" division? Well, that is possible, with unchecked_div. Most people do not need unchecked_div. If you think you do but you haven't benchmarked yet, you do not. It doesn't get any simpler than that.

This attitude is why modern software is dogshit slow. People make this "if you haven't benchmarked, it doesn't matter" argument thousands of times and the result is that every program I run is slower than molasses in Siberia. I don't care about "safety" at the expense of performance.

The optimal control framework Casadi has had the ability to compute sparse jacobians and hessians for a long time (maybe a decade?), which come up all the time in trajectory optimization.This not only provides massive speed ups in both the differentiation and linear solver time, but also greatly reduces the memory requirements. If this catches on in machine learning, it will be interesting to see if we can finally move past first order optimization methods.